Field notes
Working notes from the harness
Decks, diagrams and write-ups distilled from real engagements — the patterns we keep reaching for, and the ones we've had to retire.
Field note · 01
The 7 levels of Claude Code & RAG
A 1920×1080 deck distilling Chase AI's framework — from naive lookups to agentic retrieval — through the lens of how we build retrieval that actually earns its place in production.
rag · agents · claude code
Field note · 02
Four levels of agentic AI
A nine-slide briefing on the four levels of agentic AI — from a chatbot in a tab to a production-grade agentic system, and the harness that makes the leap possible.
agents · harness · operations
Field note · 03
Fine-tuned models, engineered to keep working
A four-slide engineering view of fine-tuned models — the adapter, the evaluation harness, and the retraining loop. Why a one-off checkpoint decays, and how the loop holds the line.
fine-tuning · evals · lora